An effective data transformation approach for privacy preserving similarity measurement

Abstract

Data similarity measurement is an important direction for data mining research. This paper is concentrated on the issue of protecting the underlying attribute values when sharing data for the similarity of objects measurement and proposes a simple data transformation method: Isometric-Based Transformation (IBT). IBT selects the attribute pairs and then distorts them with Isometric Transformation. In the process of transformation, the goal is to find the proper angle ranges to satisfy the least privacy preserving requirement and then randomly choose one angle in this interval. The experiment demonstrates that the method can distort attribute values, preserve privacy information and guarantee valid similarity measurement.